registry / @yhong91/vibetime / 0.1.50

@yhong91/vibetime@0.1.50

vibetime CLI — install AI-agent hooks (Claude Code, Codex, OpenCode, Pi) and report activity to vibetime.

AI Security Review

scanned 9d ago · by lpm-firewall-ai

The CLI can install persistent hooks into multiple AI-agent configurations. Those hooks trigger background import and upload of local AI-work activity to a hard-coded HTTP endpoint.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

User runs `vibetime install` and subsequently uses an integrated AI agent.

Impact

Exfiltrates detailed developer/agent activity metadata and file-path information to a remote service over HTTP.

Mechanism

persistent agent hooks invoke a detached telemetry uploader

Attack narrative

A user-invoked install command adds hooks to supported AI-agent configuration files. On agent events, the generated hook launches a detached runner that reads local session histories, builds activity/file-path rollups, and POSTs them to the package's default hard-coded HTTP server. The implementation does not require a token before attempting ingestion.

Rationale

This is a concrete, persistent telemetry-exfiltration chain to an unencrypted hard-coded IP endpoint. Explicit installation and metadata-only aggregation reduce ambiguity, but do not eliminate the material data-exfiltration behavior.

Evidence

package.jsonbin/vibetime.mjs~/.claude/settings.json~/.codex/hooks.json~/.vibetime/config.json~/.vibetime/machine-id

Network endpoints1

121.196.224.82:3001/v3/agent/ingest

@yhong91/vibetime@0.1.50

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code